]> git.rkrishnan.org Git - tahoe-lafs/tahoe-lafs.git/commitdiff
Add src/allmydata/util/verlib.py, which is a copy of verlib from https://bitbucket...
authordavid-sarah <david-sarah@jacaranda.org>
Fri, 21 Jan 2011 06:04:49 +0000 (22:04 -0800)
committerdavid-sarah <david-sarah@jacaranda.org>
Fri, 21 Jan 2011 06:04:49 +0000 (22:04 -0800)
src/allmydata/util/verlib.py [new file with mode: 0644]

diff --git a/src/allmydata/util/verlib.py b/src/allmydata/util/verlib.py
new file mode 100644 (file)
index 0000000..8c5e0ce
--- /dev/null
@@ -0,0 +1,327 @@
+"""
+"Rational" version definition and parsing for DistutilsVersionFight
+discussion at PyCon 2009.
+"""
+
+
+import re
+
+class IrrationalVersionError(Exception):
+    """This is an irrational version."""
+    pass
+
+class HugeMajorVersionNumError(IrrationalVersionError):
+    """An irrational version because the major version number is huge
+    (often because a year or date was used).
+
+    See `error_on_huge_major_num` option in `NormalizedVersion` for details.
+    This guard can be disabled by setting that option False.
+    """
+    pass
+
+# A marker used in the second and third parts of the `parts` tuple, for
+# versions that don't have those segments, to sort properly. An example
+# of versions in sort order ('highest' last):
+#   1.0b1                 ((1,0), ('b',1), ('f',))
+#   1.0.dev345            ((1,0), ('f',),  ('dev', 345))
+#   1.0                   ((1,0), ('f',),  ('f',))
+#   1.0.post256.dev345    ((1,0), ('f',),  ('f', 'post', 256, 'dev', 345))
+#   1.0.post345           ((1,0), ('f',),  ('f', 'post', 345, 'f'))
+#                                   ^        ^                 ^
+#   'b' < 'f' ---------------------/         |                 |
+#                                            |                 |
+#   'dev' < 'f' < 'post' -------------------/                  |
+#                                                              |
+#   'dev' < 'f' ----------------------------------------------/
+# Other letters would do, but 'f' for 'final' is kind of nice.
+FINAL_MARKER = ('f',)
+
+VERSION_RE = re.compile(r'''
+    ^
+    (?P<version>\d+\.\d+)          # minimum 'N.N'
+    (?P<extraversion>(?:\.\d+)*)   # any number of extra '.N' segments
+    (?:
+        (?P<prerel>[abc]|rc)       # 'a'=alpha, 'b'=beta, 'c'=release candidate
+                                   # 'rc'= alias for release candidate
+        (?P<prerelversion>\d+(?:\.\d+)*)
+    )?
+    (?P<postdev>(\.post(?P<post>\d+))?(\.dev(?P<dev>\d+))?)?
+    $''', re.VERBOSE)
+
+class NormalizedVersion(object):
+    """A rational version.
+
+    Good:
+        1.2         # equivalent to "1.2.0"
+        1.2.0
+        1.2a1
+        1.2.3a2
+        1.2.3b1
+        1.2.3c1
+        1.2.3.4
+        TODO: fill this out
+
+    Bad:
+        1           # mininum two numbers
+        1.2a        # release level must have a release serial
+        1.2.3b
+    """
+    def __init__(self, s, error_on_huge_major_num=True):
+        """Create a NormalizedVersion instance from a version string.
+
+        @param s {str} The version string.
+        @param error_on_huge_major_num {bool} Whether to consider an
+            apparent use of a year or full date as the major version number
+            an error. Default True. One of the observed patterns on PyPI before
+            the introduction of `NormalizedVersion` was version numbers like this:
+                2009.01.03
+                20040603
+                2005.01
+            This guard is here to strongly encourage the package author to
+            use an alternate version, because a release deployed into PyPI
+            and, e.g. downstream Linux package managers, will forever remove
+            the possibility of using a version number like "1.0" (i.e.
+            where the major number is less than that huge major number).
+        """
+        self._parse(s, error_on_huge_major_num)
+
+    @classmethod
+    def from_parts(cls, version, prerelease=FINAL_MARKER,
+                   devpost=FINAL_MARKER):
+        return cls(cls.parts_to_str((version, prerelease, devpost)))
+
+    def _parse(self, s, error_on_huge_major_num=True):
+        """Parses a string version into parts."""
+        match = VERSION_RE.search(s)
+        if not match:
+            raise IrrationalVersionError(s)
+
+        groups = match.groupdict()
+        parts = []
+
+        # main version
+        block = self._parse_numdots(groups['version'], s, False, 2)
+        extraversion = groups.get('extraversion')
+        if extraversion not in ('', None):
+            block += self._parse_numdots(extraversion[1:], s)
+        parts.append(tuple(block))
+
+        # prerelease
+        prerel = groups.get('prerel')
+        if prerel is not None:
+            block = [prerel]
+            block += self._parse_numdots(groups.get('prerelversion'), s,
+                                         pad_zeros_length=1)
+            parts.append(tuple(block))
+        else:
+            parts.append(FINAL_MARKER)
+
+        # postdev
+        if groups.get('postdev'):
+            post = groups.get('post')
+            dev = groups.get('dev')
+            postdev = []
+            if post is not None:
+                postdev.extend([FINAL_MARKER[0], 'post', int(post)])
+                if dev is None:
+                    postdev.append(FINAL_MARKER[0])
+            if dev is not None:
+                postdev.extend(['dev', int(dev)])
+            parts.append(tuple(postdev))
+        else:
+            parts.append(FINAL_MARKER)
+        self.parts = tuple(parts)
+        if error_on_huge_major_num and self.parts[0][0] > 1980:
+            raise HugeMajorVersionNumError("huge major version number, %r, "
+                "which might cause future problems: %r" % (self.parts[0][0], s))
+
+    def _parse_numdots(self, s, full_ver_str, drop_trailing_zeros=True,
+                       pad_zeros_length=0):
+        """Parse 'N.N.N' sequences, return a list of ints.
+
+        @param s {str} 'N.N.N..." sequence to be parsed
+        @param full_ver_str {str} The full version string from which this
+            comes. Used for error strings.
+        @param drop_trailing_zeros {bool} Whether to drop trailing zeros
+            from the returned list. Default True.
+        @param pad_zeros_length {int} The length to which to pad the
+            returned list with zeros, if necessary. Default 0.
+        """
+        nums = []
+        for n in s.split("."):
+            if len(n) > 1 and n[0] == '0':
+                raise IrrationalVersionError("cannot have leading zero in "
+                    "version number segment: '%s' in %r" % (n, full_ver_str))
+            nums.append(int(n))
+        if drop_trailing_zeros:
+            while nums and nums[-1] == 0:
+                nums.pop()
+        while len(nums) < pad_zeros_length:
+            nums.append(0)
+        return nums
+
+    def __str__(self):
+        return self.parts_to_str(self.parts)
+
+    @classmethod
+    def parts_to_str(cls, parts):
+        """Transforms a version expressed in tuple into its string
+        representation."""
+        # XXX This doesn't check for invalid tuples
+        main, prerel, postdev = parts
+        s = '.'.join(str(v) for v in main)
+        if prerel is not FINAL_MARKER:
+            s += prerel[0]
+            s += '.'.join(str(v) for v in prerel[1:])
+        if postdev and postdev is not FINAL_MARKER:
+            if postdev[0] == 'f':
+                postdev = postdev[1:]
+            i = 0
+            while i < len(postdev):
+                if i % 2 == 0:
+                    s += '.'
+                s += str(postdev[i])
+                i += 1
+        return s
+
+    def __repr__(self):
+        return "%s('%s')" % (self.__class__.__name__, self)
+
+    def _cannot_compare(self, other):
+        raise TypeError("cannot compare %s and %s"
+                % (type(self).__name__, type(other).__name__))
+
+    def __eq__(self, other):
+        if not isinstance(other, NormalizedVersion):
+            self._cannot_compare(other)
+        return self.parts == other.parts
+
+    def __lt__(self, other):
+        if not isinstance(other, NormalizedVersion):
+            self._cannot_compare(other)
+        return self.parts < other.parts
+
+    def __ne__(self, other):
+        return not self.__eq__(other)
+
+    def __gt__(self, other):
+        return not (self.__lt__(other) or self.__eq__(other))
+
+    def __le__(self, other):
+        return self.__eq__(other) or self.__lt__(other)
+
+    def __ge__(self, other):
+        return self.__eq__(other) or self.__gt__(other)
+
+def suggest_normalized_version(s):
+    """Suggest a normalized version close to the given version string.
+
+    If you have a version string that isn't rational (i.e. NormalizedVersion
+    doesn't like it) then you might be able to get an equivalent (or close)
+    rational version from this function.
+
+    This does a number of simple normalizations to the given string, based
+    on observation of versions currently in use on PyPI. Given a dump of
+    those version during PyCon 2009, 4287 of them:
+    - 2312 (53.93%) match NormalizedVersion without change
+    - with the automatic suggestion
+    - 3474 (81.04%) match when using this suggestion method
+
+    @param s {str} An irrational version string.
+    @returns A rational version string, or None, if couldn't determine one.
+    """
+    try:
+        NormalizedVersion(s)
+        return s   # already rational
+    except IrrationalVersionError:
+        pass
+
+    rs = s.lower()
+
+    # part of this could use maketrans
+    for orig, repl in (('-alpha', 'a'), ('-beta', 'b'), ('alpha', 'a'),
+                       ('beta', 'b'), ('rc', 'c'), ('-final', ''),
+                       ('-pre', 'c'),
+                       ('-release', ''), ('.release', ''), ('-stable', ''),
+                       ('+', '.'), ('_', '.'), (' ', ''), ('.final', ''),
+                       ('final', '')):
+        rs = rs.replace(orig, repl)
+
+    # if something ends with dev or pre, we add a 0
+    rs = re.sub(r"pre$", r"pre0", rs)
+    rs = re.sub(r"dev$", r"dev0", rs)
+
+    # if we have something like "b-2" or "a.2" at the end of the
+    # version, that is pobably beta, alpha, etc
+    # let's remove the dash or dot
+    rs = re.sub(r"([abc|rc])[\-\.](\d+)$", r"\1\2", rs)
+
+    # 1.0-dev-r371 -> 1.0.dev371
+    # 0.1-dev-r79 -> 0.1.dev79
+    rs = re.sub(r"[\-\.](dev)[\-\.]?r?(\d+)$", r".\1\2", rs)
+
+    # Clean: 2.0.a.3, 2.0.b1, 0.9.0~c1
+    rs = re.sub(r"[.~]?([abc])\.?", r"\1", rs)
+
+    # Clean: v0.3, v1.0
+    if rs.startswith('v'):
+        rs = rs[1:]
+
+    # Clean leading '0's on numbers.
+    #TODO: unintended side-effect on, e.g., "2003.05.09"
+    # PyPI stats: 77 (~2%) better
+    rs = re.sub(r"\b0+(\d+)(?!\d)", r"\1", rs)
+
+    # Clean a/b/c with no version. E.g. "1.0a" -> "1.0a0". Setuptools infers
+    # zero.
+    # PyPI stats: 245 (7.56%) better
+    rs = re.sub(r"(\d+[abc])$", r"\g<1>0", rs)
+
+    # the 'dev-rNNN' tag is a dev tag
+    rs = re.sub(r"\.?(dev-r|dev\.r)\.?(\d+)$", r".dev\2", rs)
+
+    # clean the - when used as a pre delimiter
+    rs = re.sub(r"-(a|b|c)(\d+)$", r"\1\2", rs)
+
+    # a terminal "dev" or "devel" can be changed into ".dev0"
+    rs = re.sub(r"[\.\-](dev|devel)$", r".dev0", rs)
+
+    # a terminal "dev" can be changed into ".dev0"
+    rs = re.sub(r"(?![\.\-])dev$", r".dev0", rs)
+
+    # a terminal "final" or "stable" can be removed
+    rs = re.sub(r"(final|stable)$", "", rs)
+
+    # The 'r' and the '-' tags are post release tags
+    #   0.4a1.r10       ->  0.4a1.post10
+    #   0.9.33-17222    ->  0.9.33.post17222
+    #   0.9.33-r17222   ->  0.9.33.post17222
+    rs = re.sub(r"\.?(r|-|-r)\.?(\d+)$", r".post\2", rs)
+
+    # Clean 'r' instead of 'dev' usage:
+    #   0.9.33+r17222   ->  0.9.33.dev17222
+    #   1.0dev123       ->  1.0.dev123
+    #   1.0.git123      ->  1.0.dev123
+    #   1.0.bzr123      ->  1.0.dev123
+    #   0.1a0dev.123    ->  0.1a0.dev123
+    # PyPI stats:  ~150 (~4%) better
+    rs = re.sub(r"\.?(dev|git|bzr)\.?(\d+)$", r".dev\2", rs)
+
+    # Clean '.pre' (normalized from '-pre' above) instead of 'c' usage:
+    #   0.2.pre1        ->  0.2c1
+    #   0.2-c1         ->  0.2c1
+    #   1.0preview123   ->  1.0c123
+    # PyPI stats: ~21 (0.62%) better
+    rs = re.sub(r"\.?(pre|preview|-c)(\d+)$", r"c\g<2>", rs)
+
+
+    # Tcl/Tk uses "px" for their post release markers
+    rs = re.sub(r"p(\d+)$", r".post\1", rs)
+
+    try:
+        NormalizedVersion(rs)
+        return rs   # already rational
+    except IrrationalVersionError:
+        pass
+    return None
+